Withdrawal Experience

[Nireshan]

Hi Guys

I have a question on calculating withdrawal experience.

Let's assume we are the end of 2012 and are looking at data for the last 4 years (2009 onwards)

Why is that when I calculate the withdrawal rate for year 1, I do not take into account those policies that were sold in 2009 and who lapsed in their first year. (between 2009 and 2010)

The way the notes are worded suggests that you would not consider these policies to calc year 1 rate.

Nireshan

 

[jollyfakey]

Which notes implied this? I am not sure about this assertion though, But i think policies sold and lapsed in 2009 are included in calculating first year persistence rate, Unless your calculations start from 31st december 2009.

Except you are looking at the exposure. This you can over come by using exposed-to-risk period in calculating withdrawal rates.

 

[Nireshan]

Chapter 37, page 8,the third last paragraph reads:

"The number of contracts issued in the company's last financial year is divided into the number that survive in force until the first policy anniversary to give a first year persistancy rate. Withdrawal rate = 1 minus persistancy rate"


Based on my example, how do policies sold in 2009 feature here?

 

[jollyfakey]

Hello Nireshan,

Except i am missing your point completely, i think the example you quoted answers your question. If 2009, for example, is the first year and we are at end of 2009 then;

"The number of contracts issued in the company's last financial year" --refers to 2009 since we are the end of the year,

and

"the number that survive in force until the first policy anniversary" --also refers to 2009 survivors

 

[morrisja]

I believe what Nireshan is saying is more like this:

We have New Business in years 2009, 2010, 2011, 2012.

Should our year one lapse rate account for the New business sold just in the last year (i.e. any lapses in 2012 NB) or should we also use data from previous years. So we have NB in 2009, if any of it lapsed within a year of issue (between 2009/2010) should we include this data in our calculation of the year 1 lapse rate? And so on for 2010 and 2011 NB.

Thoughts on this:

Typically other investigations will use several years of data, so our mortality investigations would use maybe the last 5 years of data.

As far as the answer goes I'm not 100% sure. The pro of extra data may be outweighed by the likely heterogeneity.

Withdrawal experience is highly variable and has a lot more time dependent factors than say, mortality experience. Withdrawals depend a lot on the economic environment and investment performance. It's quite possible that data from 4 years ago would not be appropriate for estimating future withdrawal experience.

This may come down to a judgement call on the specific circumstances of your data and environment.

I'd be interested to hear other opinions though.

 

[Helloall]

I too was wondering how withdrawal rates are actually calculated.

If i am conducting a withdrawal experience analysis, i want some data (preferably last year and maybe a few more).

If i say I had the 2018,2019,2020,2021 Year end data, then how do i make estimates for withdrawal?

I assume the first thing i do is split the 2021 YE data into a homogeneous group i am interested in, lets say distribution channel. So i take only insurance intermediary data out of this.

I then want to calculate withdrawal rates. To determine the policies first year withdrawal rate, what do i do?

Do i say, the number of policies issued in 2020 is x, the number of these in force at 2021 is y. Hence first year withdrawal rate is 1 - Y/x.
Then do i say, the number of policies issued in 2019 is a, the number of these in force at 2021 is b. Hence second year withdrawal rate is 1 - b/a?

and just repeat?

Do i then compare this analysis to the same analysis done at 2020, 2019? Or is there something I am missing.

P.s. Mark Wilder, thanks for the help on the previous question I asked earlier in the week.

 

[Mark Willder]

I too was wondering how withdrawal rates are actually calculated.

If i am conducting a withdrawal experience analysis, i want some data (preferably last year and maybe a few more).

If i say I had the 2018,2019,2020,2021 Year end data, then how do i make estimates for withdrawal?

I assume the first thing i do is split the 2021 YE data into a homogeneous group i am interested in, lets say distribution channel. So i take only insurance intermediary data out of this.

I then want to calculate withdrawal rates. To determine the policies first year withdrawal rate, what do i do?

Do i say, the number of policies issued in 2020 is x, the number of these in force at 2021 is y. Hence first year withdrawal rate is 1 - Y/x.
Then do i say, the number of policies issued in 2019 is a, the number of these in force at 2021 is b. Hence second year withdrawal rate is 1 - b/a?

and just repeat?

Do i then compare this analysis to the same analysis done at 2020, 2019? Or is there something I am missing.

P.s. Mark Wilder, thanks for the help on the previous question I asked earlier in the week.

Your initial comments look spot on, but I would change the second year formula.

For the first year you have had x-y withdrawals, so the withdrawal rate is (x-y)/x = 1 -y/x as you suggest.

For your second year withdrawal rate, you've included two years worth of withdrawals here rather than one. Let's say that the policies issued in 2019 is A, those remaining in 2020 is B and then we have C in 2021. Then I'd say the second year rate is then (B-C)/B.

Best wishes

Mark